" Multi-Period Resource Allocation at System Edges - Capacity Management in a Multi-Provider Multi-Service Internet "

Oliver Heckmann, Jens Schmitt

Industrial Process and System Communications,
Darmstadt University of Technology, Germany


Providing guaranteed QoS, be it statistical or deterministic, necessarily requires allocation of scarce resources. This might happen on a session or on an aggregate basis, nevertheless, it is conceivable that at least at system edges scarcity of resources, exposed in the form of non-negligible (virtual) costs, will prevail to necessitate explicit allocation of resources as opposed to pure overdimensioning.

An example of this logic is constituted by the Differentiated Services (DiffServ) architecture which is largely based on explicit bilateral Service Level Agreements (SLA) between peering providers. Often such resource allocation decisions are done on a multi-period basis because resource allocation decisions at a certain point in time may depend on earlier decisions and thus it can turn out sub-optimal to look at decisions in an isolated fashion.

Therefore, in this paper, we investigate a fairly large and diverse set of (network) QoS problems all of which deal with the problem of multi-period resource allocation at system edges. We devise a taxonomy for the classification of these problems and introduce a common mathematical framework under which these problems can be tackled.

The ultimate goal of our work is to strive for solution techniques towards the generalized class of problems such that these are applicable in a number of scenarios which have so far not been regarded in an integrated fashion.